StatsD Line Parser

How to use

1. 1Paste one or more StatsD or DogStatsD lines into the box, one metric per line, or load a sample such as Core types or DogStatsD tags.
2. 2Read each metric card: the decoded name, the metric type with its meaning, the value or set member, the sample rate, and any tags.
3. 3Watch the type chips and the delta marker on gauges that use a leading + or - to change the value rather than set it.
4. 4Check the errors panel for lines that would be dropped at a collector, and the notes panel for softer warnings like an unusual sample rate.
5. 5Use Copy JSON or Copy NDJSON to export every parsed metric as structured data.
6. 6Switch to Build a line to assemble a valid line from a form, with the type, sample rate, and tag block written in the correct order.

About this tool

StatsD Line Parser reads the metric lines that StatsD speaks: the short, pipe-delimited datagrams that application code, Telegraf, the Datadog Agent, and the original Etsy StatsD daemon send over UDP or TCP. The grammar fits on one line but the details matter, so this tool implements the documented rules. A line is name:value|type, with an optional sample rate and an optional tag block: name:value|type|@rate|#tag:value,tag2. The parser splits the name at the first colon, reads the value section, then walks the pipe-separated fields for the type, the @rate, and the #tags in whatever order they appear. Every metric type is recognized and labelled with what it means: c is a counter that adds to a bucket each flush, g is a gauge that holds a value, ms is a timer in milliseconds, s is a set that counts unique members, and the two DogStatsD additions, h for a histogram and d for a distribution, round out the set. Gauges are handled correctly: a leading + or - is not part of the number, it is a delta that changes the gauge by that amount instead of setting it, so queue.depth:+4|g and queue.depth:-2|g are shown as deltas, not as the values 4 and -2. Value packing is supported too, the DogStatsD shorthand that sends several readings on one line as 1:2:32, so a packed distribution or histogram is split into its individual values. The sample rate is parsed and checked against the usual (0, 1] range, which is how a client tells the server that it sent one in ten or one in two, and the server scales the count back up. Tags are parsed into key and value pairs, with a bare key allowed when a tag carries no value. It is a linter as well as a reader. It flags the mistakes that silently drop metrics at the collector: a line with no colon or no type, a value that is not a number, a type that is not one of the six codes, a sample rate that is not a number or sits outside the normal range, an empty or stray field from a doubled pipe, and a tag block that is empty. Each parsed line is laid out as a card with a colored type chip, the value or set member, the sample rate, and the tag list, and the whole batch exports as JSON or NDJSON with the type spelled out, the value or packed values, a delta flag for gauges, the sample rate, and the tags as an object, ready for a script, a test fixture, or a spreadsheet. A build mode goes the other way: pick a type, enter a value, choose set-or-delta for a gauge, add a sample rate and tags, and the tool assembles one valid line with the pipe fields and tag block in the right order. One honest limit: this reads and writes the StatsD text, it is not a StatsD server. It does not listen on a socket, receive from a client, aggregate, or flush. Everything runs locally in your browser, so the lines you paste or build are never uploaded or logged.

Free to use. Works in your browser. No signup, no login.